Interval Estimation for Minimal Clinically Important Difference and its Classification Error via a Bootstrap Scheme

Area of Science:

• Clinical research methodology
• Statistical analysis in healthcare
• Patient-reported outcomes

Background:

• Clinical researchers increasingly use minimal clinically important difference (MCID) over statistical significance for treatment effectiveness.
• Patient-reported outcome data is more accessible, enhancing clinical relevance.
• Current MCID determination methods, based on diagnostic measurements and large margin classification, provide only point estimations.

Purpose of the Study:

• To introduce a novel bootstrap approach for estimating MCID.
• To provide interval estimations for MCID and its classification error.
• To develop an associated accuracy measure for performance assessment of MCID estimation.

Main Methods:

• An m-out-of-n bootstrap approach is proposed.
• The method provides interval estimations for MCID.
• Classification error and accuracy measures are derived from the bootstrap intervals.

Main Results:

• The proposed bootstrap method yields interval estimations for MCID.
• The method allows for the assessment of MCID estimation performance through classification error and accuracy.
• Simulation studies demonstrate the advantages of the proposed method.

Conclusions:

• The m-out-of-n bootstrap approach offers a robust method for interval estimation of MCID.
• This technique enhances the performance assessment of MCID in clinical trials.
• The method was illustrated using data from the chondral lesions and meniscus procedures (ChAMP) trial.